Pranava Kailash Subramaniam Prema
I build Python and SQL ETL pipelines for regulatory data, schedule data workflows with Apache Airflow, and work with AWS ingestion and storage.
I build data pipelines that turn messy inputs into datasets people can use. My background includes commercial ETL work and an MSc in Data Science from the University of Surrey.
At Fleet Street Research, I worked on Python and SQL pipelines for regulatory data: collecting records, transforming them across MySQL and SQLite, and checking for missing data and schema changes. I also built Power BI semantic models and DAX measures for risk and compliance reporting.
My independent projects use Apache Airflow for scheduled data workflows and AWS Kinesis, EMR, and S3 for weather-data ingestion and processing. Machine learning remains part of my background, but data engineering is where I want to focus.
Ingest records, clean them, and keep transformations traceable for the people who use the results.
Catch schema changes, missing records, and processing errors before they affect reporting.
Use schedules, retries, and clear documentation so a pipeline can run beyond a single demo.
Oct 2022 — Jan 2024
Fleet Street Research · Freelance · UK
Feb 2022 — Apr 2022
Yoshops.com · India
Mar 2021 — Jun 2021
Forsk Coding School · India
I put the data engineering work first: ingestion, orchestration, storage, and validation. The ML and product projects show where that work has taken me next.
01
A scheduled workflow that collects NVIDIA price history and financial news for a dashboard of indicators and sentiment.
02
A containerised pipeline that extracts weather observations from OpenWeather, transforms the API response, and loads it into PostgreSQL.
03
Data engineering · AWS project
An independent project connecting streaming weather-sensor ingestion, batch processing, and cloud storage.
04
A cybersecurity named-entity recognition system for finding malware, vulnerabilities, threat actors, systems, and organisations in text.
05
A Chrome extension that rewrites prompts inside ChatGPT, Gemini, Claude, and Perplexity using Chrome’s on-device Gemini Nano model.
These are tools I have used in client work or independent projects. Each area points back to a pipeline, report, or system on this page.
Python · SQL · Apache Airflow · MySQL · SQLite
Commercial compliance-data pipelines and scheduled financial and weather workflows.
AWS Kinesis · Amazon EMR · Amazon S3 · Docker
Independent weather-sensor pipeline from streaming ingestion to batch processing and storage.
Power BI · DAX · pandas · Data validation
Fleet Street compliance reporting, semantic models, and checks for schema drift and missing records.
scikit-learn · Hugging Face Transformers · FastAPI
Cybersecurity entity recognition, evaluation, and API-based inference.
Longer explanations of the systems, experiments, and decisions behind my work.
Follow on MediumDeepSeek-V3 splits every layer into 256 experts and fires only 8. The parameter count is the marketing number. The router is the product.
A weekend hobby project and unreleased enterprise security software are betting on the same architecture: never trust one model alone.
From a single ChatGPT button to four platforms, seven versions, and a published Chrome Web Store extension. Here’s what I actually learned
I am based in Guildford and open to UK data engineering roles with Skilled Worker sponsorship.